GAM-of-1 Analyses: A Tutorial on Leveraging Generalized Additive Models for Rich and Complex N-of-1 Data

Average effects models such as regression analyses smooth over the individual differences naturally present in populations. This prevents researchers, practitioners, and other stakeholders from understanding individual-level insights about patients, students, and other subjects of interest. The n-of-1 design has become increasingly more popular in recent years for modeling fine-grained data from experimental interventions and provides deep insight into individual-level outcomes. Because n-of-1 data often exhibit nonlinearity, confounding, and repeated measures, it would be beneficial to use methods which account for this complexity. To the best of our knowledge, there does not yet exist a framework for dealing with these common problems in n-of-1 designs. Our paper illustrates how to combine the n-of-1 study design with generalized additive models (GAMs) to account for this problem in what can be called a “GAM-of-1” design. Using basic demonstrations with simulations and a real-world case study, we show how GAM-of-1 studies can provide deep insights into individuals of interest, particularly if they have rare conditions or extreme responses that are difficult to account for in standard models.